Instance data for crane scheduling for end-of-aisle picking

The following instances are from the working paper entitled Crane scheduling for end-of-aisle picking: Complexity and efficient solutions based on the vehicle routing problem by Nils Boysen, Konrad Stephan, and Simon Emde.

The dataset contains 760 randomly generated instances with varying numbers of requests and buffer sizes. Each line in the attached text file stands for one instance. Columns are separated by tab stops. The individual columns are:

ID - unique running number

n - number of requests in set S

b - buffer size, number of requests in set S_0

S_0 - set of request locations initially in the buffer

S - set of request locations

FCFS_SOLUTION - sequence vector (solution) in a first come, first served schedule

FCFS_TIME - CPU time (milliseconds) for finding the first come, first served solution

FCFS_OBJECTIVE - objective value (tenths of a second of empty crane travel) of the first come, first served solution

MBH_SOLUTION - sequence vector (solution) found by the matching-based heuristic

MBH_TIME - CPU time (milliseconds) taken by the matching-based heuristic

MBH_OBJECTIVE - objective value (tenths of a second of empty crane travel) of the solution found by the matching-based heuristic

VRP_SOLUTION - sequence vector (solution) of the VRP* (via HGA-CVRP) schedule

VRP_TIME - CPU time (milliseconds) for finding the VRP* solution

VRP_OBJECTIVE - objective value (tenths of a second of empty crane travel) of the VRP* solution

SA_SOLUTION - sequence vector (solution) found by the simulated annealing heuristic

SA_TIME - CPU time (milliseconds) taken by the simulated annealing heuristic

SA_OBJECTIVE - objective value (tenths of a second of empty crane travel) of the solution found by the simulated annealing heuristic

Note that location data is in the format of two-dimensional coordinates denoting the travel time on the x- and y-axes in tenths of a second from coordinate (0,0). Hence coordinate "(600, 300)" implies that the corresponding slot is 60 seconds of horizontal travel and 30 seconds of vertical travel away from the origin. Objective values are also scaled to tenths of a second. The CPU times are given in milliseconds. Results for the simulated annealing heuristic are only available for instances with n=60 requests.